Creating DLL's for Lisp-Stat on Windows

B. Narasimhan
Department of Statistics
Stanford University
Stanford, CA 94305

Revision: 1.2 of 1998/05/19

Abstract

This is a short document that I wrote for myself (and Forrest Young) with Luke Tierney's help on how to create DLL's for use with Lisp-Stat.

Introduction

[*]

This document is a short introduction to creating DLL's on Windows. I expect it to evolve into a document covering how to create dynamically loadable modules on all platforms for Lisp-Stat. I use Microsoft Visual C++ version 5.0 for all the examples below.

Example

[*]

Consider an almost identical example to the one described in Tierney's book[cite tier:1990].

<C subroutine>=
void foo (int *n, double *x, double *sum) {
  int i;

  *sum = 0.0;
  for (i = 0; i < *n; i++) {
     *sum += x[i];
  }
}

Defines foo (links are to index).

Suppose we wish to create a DLL in Windows to load this program in foo.c, how can we do it?

Download the Lisp-Stat sources onto your machine from the University of Minnesota, Department of Statistics ftp server. You need the actual sources, not just the binaries for Windows.
The Lisp-Stat sources are supposed to contain a file called wxls32.def in the subdirectory msdos for creating DLL's but this is not the case at present. Tierney will be fixing this in the future, so for now, you need to use the file he supplied me.
Move the file wxls32.def suppled here to the Lisp-Stat source subdirectory msdos. You only need to do this once regardless of how many DLL's you create.
Create a subdirectory for creating windows code, say win. Move the file foo.c In this subdirectory. Edit the C program and prepend each function you want to call in Lisp-Stat with a declaration that will ensure that it is exported.
```
<Export declaration>=
__declspec(dllexport) void foo(int *n, double *x, double *sum);
```

Copy the makefile.vc present in the Lisp-Stat sources in the subdirectory Extras/sockets to the current directory and edit it to suit your needs. For example, here is the one I use for the above example. Note how you have to quote directory names containing spaces.

<Makefile for Visual C++ 5.x>=
# Change the next line to point to where your xlispstat sources are.
XLSDIR = d:\xls-3-52.5
WXLSDIR = $(XLSDIR)\msdos
XLSLIB = vcwxls32.lib

CC=cl.exe
LINK32=link.exe
LIB32=lib.exe

DLL_CFLAGS=/nologo /MT /W3 /GX /O2 /D "WIN32" /D "NDEBUG" /D "_Windows" /FD \
  /I $(WXLSDIR) /I $(XLSDIR) /I "c:\Program Files\DevStudio\Vc\include" \
  /D far=
DLL_LDFLAGS=/LIBPATH:"c:\Program Files\DevStudio\Vc\lib" $(STDLIBS) \
 /nologo /subsystem:windows /dll /incremental:no /machine:I386

STDLIBS=kernel32.lib user32.lib gdi32.lib winspool.lib comdlg32.lib\
 advapi32.lib shell32.lib ole32.lib oleaut32.lib uuid.lib odbc32.lib\
 odbccp32.lib wsock32.lib

.c.obj:
        $(CC) $(DLL_CFLAGS) -c $<
{$(WXLSDIR)}.c.obj:
        $(CC) $(DLL_CFLAGS) -c $<

OBJECTS = $(WXLSDIR)\dllstub.obj foo.obj

foo.dll : $(OBJECTS) $(XLSLIB)
    $(LINK32) <<
  $(DLL_LDFLAGS) /out:foo.dll dllstub.obj foo.obj $(XLSLIB)
 <<

vcwxls32.lib: $(WXLSDIR)/wxls32.def
   $(LIB32) /def:$(WXLSDIR)/wxls32.def /out:vcwxls32.lib

clean :
        -@erase *.obj
        -@erase *.dll
        -@erase *.exp
        -@erase *.lib
        -@erase *.exe
        -@erase *.idb

The makefile also assumes that the compiler tools will be found in your path.

Bring up an MSDOS window and in the win subdirectory make the DLL.
```
<Command to create the DLL>=
nmake foo.dll
```

That's it. Now the DLL can be used. As an example, let's try the call-cfun stuff. Note how we need the oldcfun stuff described on Tierney's project page.

<Use the DLL>=
(require "oldcfun")
(def lib (shlib::shlib-open "foo.dll"))
(def w (shlib::old-call-cfun "foo" lib 5 (float (iseq 1 5)) 0.0))

<C subroutine>: D1
<Command to create the DLL>: D1
<Export declaration>: D1
<Makefile for Visual C++ 5.x>: D1
<Use the DLL>: D1

foo: D1, U2, U3, U4, U5

Concluding Remarks

[*]

One can also use a def file instead of using the export declaration, but that is clearly more inconvenient.

List of code chunks

This list is generated automatically. The numeral is that of the first definition of the chunk.

Index

Here is a list of the identifiers used, and where they appear. Underlined entries indicate the place of definition. This index is generated automatically.

References

[1] Luke Tierney. LISP-STAT: An Object-oriented Environment for Statistical Computing and Dynamic Graphics. John Wiley &Sons (New York, Chichester), 1990.